Documentation

Everything you need to run MT-CLIP.

Requirements, dependencies, first-time setup, the full clip-generation workflow, ViralEnhance, and troubleshooting — all in one place.

Download MT-CLIP ↗ View source on GitHub ↗
00:00 · GETTING STARTED

Overview

MT-CLIP is an application that automatically turns YouTube videos into short clips using AI. Simply provide a YouTube URL, choose your preferred settings, and MT-CLIP will find interesting moments and turn them into ready-to-use clips.

MT-CLIP is designed to keep the process as simple as possible — download, setup, and start creating clips.

100% FREE — until the developer changes their mind. Release ZIPs already contain everything MT-CLIP itself needs; you only need to install a few external dependencies (Ollama, FFmpeg) before running it.
00:12 · CAPABILITIES

Features

  • AI auto clip generation from YouTube
  • Automatically finds interesting moments
  • Local AI using Ollama
  • Support for multiple AI models
  • Generate multiple clips at once
  • Configurable clip count and duration
  • Landscape and Portrait / 9:16 output
  • Automatic subject tracking for portrait videos
  • Animated, customizable subtitles
  • Built-in gallery and automatic video processing
  • ViralEnhance for trend and topic analysis
  • GPU acceleration through Vulkan
  • Can run locally without paid AI APIs

Trend sources supported by ViralEnhance:

YouTubeGoogle TrendsRedditTikTok
00:24 · HARDWARE

System Requirements

MT-CLIP can run on relatively modest hardware. However, AI processing and video processing can be resource-intensive — more RAM and a better GPU generally mean faster processing.

Minimum

OSWindows 10 / 11
CPU4 Cores
RAM8 GB
GPUOptional
StorageSSD recommended
NetworkInternet connection

Recommended

CPU6–8 Cores
RAM16 GB or more
GPUDedicated, 6 GB+ VRAM
StorageSSD

For faster processing

RAM — 8 GB is usable, 16 GB is recommended, 32 GB+ is great for long videos and heavier workloads.

GPU — not required, but highly recommended. More VRAM generally provides a better experience with larger AI models (e.g. 8 GB / 12 GB / 16 GB VRAM).

Vulkan

If your GPU supports Vulkan, you can try running Ollama with Vulkan to utilize your GPU:

set OLLAMA_VULKAN=1
ollama serve
Vulkan performance varies depending on your GPU, drivers, Ollama version, and the AI model used.
00:36 · DEPENDENCIES

What You Need

To use MT-CLIP from a release, you only need to prepare a few things.

1. MT-CLIP

Download the latest release from MT-CLIP Releases ↗ and extract the ZIP wherever you want.

You do not need to download MT-CLIP's internal components separately — the release ZIP already contains everything the app itself needs.

2. Ollama

Download and install Ollama for Windows ↗. Then open a terminal:

ollama pull qwen2.5:7b

Check it installed correctly:

ollama list

If qwen2.5:7b appears in the list, the model is ready. You're free to use another compatible model — qwen2.5:7b is just the default recommendation because it's relatively lightweight and runs well locally.

3. FFmpeg

Download FFmpeg Builds by Gyan ↗ — the Full Build is recommended. After installing, verify it:

ffmpeg -version
00:48 · FIRST RUN

Initial Setup

After preparing everything above:

  1. Download MT-CLIP from the Releases page.
  2. Extract the ZIP.
  3. Install Ollama.
  4. Download your preferred AI model.
  5. Install FFmpeg.
  6. Launch MT-CLIP.
  7. Open Settings.
  8. Check the AI configuration.
  9. Configure the FFmpeg path if required.
  10. Save your settings.

MT-CLIP is now ready to use.

01:00 · WORKFLOW

How To Use

1. Enter a YouTube URL

Open the MT-CLIP Dashboard and paste your YouTube URL:

https://www.youtube.com/watch?v=xxxxxxxxxxx

2. Configure your settings

You can configure the AI model, language, number of clips, minimum/maximum duration, subtitles, portrait mode, subtitle style, and ViralEnhance. For your first run, the default settings should be enough.

3. Generate clips

Click Create Clip. MT-CLIP will automatically run through the full pipeline:

00YouTube Video
01Processing
02AI Analysis
03Find Best Moments
04Create Clips
05Add Subtitle
06Portrait Version
07Gallery

You just need to wait until the process finishes.

01:28 · HOW THE AI DECIDES

AI Clip Selection

The AI analyzes the video content and tries to find moments that have potential as short-form content. Factors it may consider:

HookEmotional impactInteresting moments Story climaxHumorPlot twist Educational valueShareabilityContextAudience engagement

The AI also tries to avoid clips that are:

  • Too short
  • Missing context
  • Cut in the middle of a statement
  • Missing a hook
  • Too similar to other clips
Results still depend on the source video, transcript, AI model, and the settings you choose.
01:40 · OPTIONAL

ViralEnhance

ViralEnhance is optional. It gives the AI additional information about topics and trends that are currently relevant, to help it judge whether a topic or discussion is relevant to current trends.

YouTube

MT-CLIP uses the YouTube Data API v3. To get an API key: open Google Cloud Console → create a project → enable YouTube Data API v3 → create an API key → enter it in MT-CLIP Settings.

Settings
  → Trend Sources
    → YouTube API Key

TikTok

MT-CLIP uses RapidAPI to retrieve TikTok data. Register on RapidAPI → find the TikTok API being used → subscribe → copy the API key → enter it in Settings.

Settings
  → Trend Sources
    → TikTok RapidAPI Key
If no API key is provided, TikTok is simply skipped — everything else keeps working.

Google Trends

Used to retrieve information about currently popular topics. Configure the region via Trend Region (example: ID). No additional API key required.

Reddit

Can also be used as an additional source of information about trending topics. If Reddit is unavailable, MT-CLIP continues to work normally.

01:52 · CONFIGURATION

AI Providers

Provider : Ollama (default)
Model : qwen2.5:7b

MT-CLIP can also use other providers compatible with its configuration:

OllamaOpenAIOpenRouter GeminiLM StudioGroqOpenAI-compatible API

For free, local usage, Ollama is the recommended choice.

02:02 · WHY LOCAL

Local AI

Using Ollama lets MT-CLIP run AI locally. Benefits:

  • No API key required
  • No paid AI API required
  • Data can be processed locally
  • GPU acceleration available
  • Freedom to choose models
  • No dependency on cloud AI services
ollama pull qwen2.5:7b

Then in MT-CLIP:

AI Provider → Ollama
AI Model → qwen2.5:7b
02:12 · VULKAN

GPU Acceleration

A GPU is not required to run MT-CLIP. However, if your computer has a Vulkan-capable GPU, you can try using it to accelerate AI processing:

set OLLAMA_VULKAN=1
ollama serve

Then launch MT-CLIP as usual.

Vulkan results and performance gains vary by hardware and drivers. A GPU with more VRAM also makes it easier to run larger AI models.
02:20 · EXPECTATIONS

Processing Time

Processing time depends on video duration and resolution, the AI model, CPU, RAM, GPU/VRAM, storage, number of clips, and your processing settings. Longer videos naturally take more time.

For faster processing

RAM16 GB+
GPUDedicated GPU
VRAM8 GB+
StorageSSD recommended
02:30 · FIXES

Troubleshooting

Ollama is not connecting

Make sure Ollama is running. Try:

ollama list

Make sure the selected model is installed:

ollama pull qwen2.5:7b

FFmpeg is not found

Make sure FFmpeg is installed. Try:

ffmpeg -version

If the command is not found, make sure FFmpeg is added to PATH or configure its location through Settings.

MT-CLIP cannot generate clips

Make sure: internet connection is available, the YouTube URL is valid, Ollama is running, the AI model is installed, FFmpeg is configured, and enough storage space is available. If the problem persists, check the application logs.

02:38 · DOWNLOAD

Release

The latest version of MT-CLIP can be downloaded via GitHub Releases. Each release ZIP already contains the files required to run MT-CLIP.

You do not need to build the application from source if you only want to use the released app.
Download Latest Release ↗
02:44 · READ THIS

Disclaimer

MT-CLIP is built to help with automated video editing and content creation. Make sure you have the appropriate rights or permission to process the videos you use.

Videos downloaded from YouTube remain subject to copyright law, YouTube's Terms of Service, platform rules, and content owner rights.

The developer is not responsible for the use of MT-CLIP to distribute copyrighted content or content that violates platform rules.
02:48 · LEGAL

License

See the LICENSE file in the MT-CLIP repository ↗ for information about the project's license.

Made with too much coffee and questionable engineering decisions.